More storage treasure: KDJ11-D/S (PDP11/53?) in H9275A cage with analogue, async, SCSI, ether

[Radix]

I'm not sure how that will help with an 11/53 board, if you can't move the memory starting address and faults in all three banks of ram...

Initially configuring it as 0.5MB and replacing the two suspect RAMS with known good parts in sockets will show if it is likely just simple ram failures...

 

[AK6DN]

I'm not sure how that will help with an 11/53 board, if you can't move the memory starting address and faults in all three banks of ram...

Initially configuring it as 0.5MB and replacing the two suspect RAMS with known good parts in sockets will show if it is likely just simple ram failures...

It certainly is true that if you have no working memory you can move to cover the physical 0-8KB or so address space loading and running any memory test diagnostic is problematic.
Some systems have very basic memory test functions in ROM that could be used as a simple memory test facility.
Or in worst case use something like PDP11GUI that will do exam/deposit console commands to memory and run a memory test. Slow, but possible. Kind of like a Turing Machine.

 

[jonathanjo]

Update: I tried a few things without any change:

• Checked voltage at the board, was dropping almost 0.5V so changed PSU, good 5V measured at the chips
• All the parts seemed cool, only the AM2908 bus transceivers even warm
• Tried freezer spray nonetheless

Still stuck at this:

RAM VPC=024454 PA=17604454 00000000/177577 <> 177776​

What's puzzling me most is that ODT reports word 0 as having sticky 201, but that conflicts with finding 177577 when expecting 177776.

And previous results are similar but slightly different.

RAM VPC=024454 PA=17604454 00000000/177576 <> 177776​

​

I suppose I'll have to start hunting for some spare RAM chips and DIP sockets.

Jonathan

PS:

That all seems very strange. I'm starting to suspect something else, bad solder joints ... refresh circuitry ...

I've looked with a magnifying glass but haven't seen anything which looks funny

home grown memory diagnostic

Thanks for the info: when I get enough RAM working to run it I certainly want to try it

 

[jonathanjo]

I really don't want to take out 54 RAM chips and find actually it's something else, so while I'm waiting for some replacement RAM chips ...

I've put together a 41256 DRAM chip tester from an Atmega 328, currently running on a breadboard.

The algorithm I've using is this:

for each pause p in range 0, 10
  for each value v in 0, 1
    write every location with the value v
    pause p milliseconds
    read every location, count how many are 1
    report

I have the low order address bits as the row address, so the chip will be refreshed during the write-loop and the read-loop. The pause will be without refresh, I'll experiment with timings up to the specified 4 ms, but will be interesting to see the decay pattern. It's written in bare-metal assembler, so the timing is pretty crisp and easy to experiment with if I find the RAM chips are marginal.

This is an "early write" /RAS and /CAS during development, to see what timing is achievable easily (here it's about 660 ns cycle). Obviously it's trivial to slow things down to meet the requirements of the -15 parts in my board.


I'll post results when I have something meaningful.

Jonathan

 

[Radix]

Hi - Great work on the ram tester - but take the easy route and don't try to remove the chips intact - some DEC boards are really tricky to desolder all the pins on to remove the chip whole, a lot have a solid power and ground layer making a giant heatsink. I almost always snip the pins off at the package with sharp, flush cut side cutters and unsolder the pins one at a time to look after the board and then fit a socket... I know it's nice to see a chip test faulty - but far better to not risk the rare vintage board!

I know people will have strong opinions on this - but just my preferred method!